Backend-as-a-Service
Prototype on a backend-as-a-service: sign-in, database and files in an afternoon.
Web, SaaS & Mobile · Architectures
The same online shop, with the same parts (catalog, orders, payments, users, emails), arranged six different ways. Switch between them and watch the parts move, then follow one order through each and see what breaks on a bad day.
01 · Six architectures
Each part keeps its colour in every drawing, so you can watch it move: out of one big app into its own service, down into a function, or into someone else's platform. The order below runs on its own; pause it or click any step.
You tap Pay. The request reaches a load balancer in front of the app.
Swipe the drawing sideways to see all of it.
One order, followed
One codebase, one process, one database. Every feature is a folder, and calling another feature is a function call.
You deploy
1 app
Databases
1
Network hops
4 + 3 in-process
Good for
Watch out for
Often built with
02 · Break it
A bug in payments, a traffic spike, a deploy, a slow database. Pick an architecture and a bad day to see which parts go down, slow down or scale out, and what customers would notice.
Architecture
What happens
all downMonolith · payments code crashes
What customers see
The whole shop is down, even browsing: everything ran in the process that crashed.
Why
In a monolith, a crash anywhere takes everything in that process with it. Running several copies helps only until each copy hits the same bug.
| all 24 outcomes | Payments code crashes | 10× traffic on product pages | Ship a change to Orders | The database slows down |
|---|---|---|---|---|
| Monolith | ||||
| Modular monolith | ||||
| Microservices | ||||
| Serverless | ||||
| Static + APIs | ||||
| Backend-as-a-Service |
A teaching model, not a benchmark: real outcomes depend on how each system is built. Safeguards (several copies, timeouts, queues, caches) change many of these cells, and each architecture has its own set.
03 · How products grow
Few products start where they end up. The usual path tracks the size of the team more than the size of the traffic.
Prototype on a backend-as-a-service: sign-in, database and files in an afternoon.
The idea works and the logic grows: one app and one database, on a platform like Cloud Run or Render.
Teams collide in one codebase: put walls between modules before splitting anything.
Parts with their own team and their own load move out into services, one at a time.
Any time, alongside the rest
Static + APIs
Marketing pages, docs and shop pages built ahead and served from a CDN.
Serverless
Spiky or scheduled jobs (image resizing, reports, webhooks) as functions beside the rest.
A common path, not a ladder everyone must climb. Plenty of successful products stay a well-kept monolith or modular monolith for good; moving right only pays off once the problems on the left actually hurt.
04 · Which one fits
Answer what you know about your product and team. The table underneath shows the trade-offs behind the answer.
What are you building first?
How many developers will work on it?
What does traffic look like?
How much control do you need over the backend and data?
Start here
Answer a question or two and a starting point appears, with how every option scored.
| rule of thumb, 5 = best | Monolith | Modular monolith | Microservices | Serverless | Static + APIs | Backend-as-a-Service |
|---|---|---|---|---|---|---|
| Fast to a first version | ||||||
| Little to run and watch | ||||||
| Scales only the busy part | ||||||
| A bug stays contained | ||||||
| Cheap while traffic is tiny | ||||||
| Many teams ship independently | ||||||
| Free to change providers | ||||||
| Easy to run on a laptop |
Rules of thumb for a typical product, not measurements: a disciplined team can make a monolith scale well or keep microservices simple. No column wins every row, which is the point.